The Representation of Stimulus Familiarity in Anterior Inferior

JOURNALOF NEUROPHYSIOLOGY
Vol. 69. No. 6. June 1993. Printed
in C’.S..i.
The Representation of Stimulus Familiarity
Temporal Cortex
LIN LI, EARL K. MILLER,
AND ROBERT DESIMONE
Luhorutorv . of‘N~rlr(~p,svcholol:v,
.
.
. National Institute of A4ental Health,
SUMMARY
AND
CONCLUSIONS
I. The inferior temporal (IT) cortex plays an important role in
both short- and long-term memory for visual patterns. Most previous studies of IT neurons have tested their responses in recency
memory tasks, which require that the memory lasts only the
length of a single behavioral trial, which may be < 1 s. To determine the role of IT neurons in longer lasting memories, we measured their responses to initially novel stimuli as the stimuli gradually became familiar to the animal.
2. Two rhesus monkeys were trained on a delayed matching to
sample (DMS) task with several intervening stimuli between the
sample and the final matching stimulus on each trial. The purpose
of the task was to ensure that the animal attended to the stimuli
and held them in memory, at least temporarily. Unlike in several
previous studies, the focus was not on within-trial
effects but
rather on the incidental memories that built up across trials as the
stimuli became familiar. Each cell was tested with a set of 20 novel
stimuli (digitized pictures of objects) that the monkey had not
seen before. These stimuli were used in a fixed order over the
course of an hour-long recording session, and the number of intervening trials between repetitions of a given sample stimulus was
varied.
3. The responses of about one-third of the cells recorded in
anterior-ventral
IT cortex declined systematically as the novel
stimuli became familiar. After six to eight repetitions, responses
reached a plateau that was -40% of the peak response. Virtually
all of these cells also showed selectivity for particular visual stimuli
and thus were not “novelty detectors” in the sense of cells that
respond to any novel stimulus. Rather, the responses of these cells
were a joint function of familiarity and specific object features
such as shape and color. A few cells showed increasing responses
with repetition over the recording session, but these changes were
accompanied by changes in baseline firing rate, suggesting that
they were caused by nonspecific effects.
4. The decrement in response with familiarity was stimulus
specific and bridged > 150 presentations of other stimuli, the maximum tested. For some cells the maximum decrement in response
occurred for those stimuli that initially elicited the largest response. There was no significant change in response to stimuli that
were already familiar.
5. The same cells that showed familiarity effects also showed
reduced responses to the matching stimuli at the end of each trial,
compared with the responses to the samples. The responses to
these matching stimuli declined with familiarity in parallel with
the decline in responses to the samples over the session. Recency
and familiarity effects appear to summate within IT cortex.
6. We examined the time course of responses to novel and
familiar stimuli. The population
of cells took 100 ms after response onset to distinguish between a novel stimulus and the same
stimulus seen once previously, possibly reflecting feedback to IT
cortex. However, after a single additional presentation, cells distinguished between novel and familiar stimuli within 10 ms of response onset, that is, by nearly the first action potential. Thus
there is virtually no time for the effect of familiarity on the initial
in Anterior Inferior
Bethesda, Marvlund
.
20892
phase of the response to be caused either by lengthy temporal
processing with IT cortex or by feedback from other structures.
7. The results support the proposal that a subpopulation
of IT
cells functions as “adaptive mnemonic filters” for both short- and
long-term memories. A high level of activation in IT cortex may
provide a feedback signal to orienting systems that the current
stimulus is new and deserving of attention. Because increased contact with a stimulus drives down activity in IT cortex, this feedback between memorv M and attentional systems may result in an
organism driven to seek out contact with new stimuli.
INTRODUCTION
Primates have a remarkable ability to retain visual information. People shown thousands of different pictures a single time each can later recognize an individual one as familiar ( Shepard 1992; Standing 1973 ) . Clearly, the neural circuitry underlying visual recognition
employs powerful
storage and retrieval mechanisms, and inferior temporal
(IT) cortex appears to be an important component of that
circuitry (Gross 1973; Mishkin 1982).
IT cortex is at the culmination
of a cortical pathway underlying the recognition of visual objects (for a review, see
Desimone and Ungerleider 1989) and is interconnected
with other structures important for memory (Aggleton et
al. 1980; Amaral and Price 1984; Baizer et al. 199 1; Desimone et al. 1980; Insausti et al. 1987a,b; Martin-Elkins
and
Horel 1992; Ungerleider and Mishkin 1982; Van Hoesen
and Pandya 1975a,b: Webster et al. 199 1). Accordingly, IT
neurons have large receptive fields and complex properties
(Desimone et al. 1984; Gross et al. 1972; Schwartz et al.
1983; Tanaka et al. 199 1). Some IT neurons appear to extract information
about the overall shape of objects,
whereas others are selective for specific stimuli such as faces
and hands. Although posterior IT cortex is most important
for visual discrimination,
anterior IT cortex is particularly
important for memory for visual form (Mishkin 1982).
The mnemonic tasks on which animals with IT lesions
are impaired include two variants of the delayed matching
to sample (DMS) paradigm, one of which requires recency
memory and the other recognition memory. In the typical
recency variant, a small set of familiar stimuli is used repeatedly. On a given trial, a sample stimulus is followed by two
test stimuli, and the animal indicates which one matches
the sample. Because both test stimuli is familiar, the task
requires that the animal judge which of the two test stimuli
was seen most recently. In the typical recognition variant,
the stimuli used are all initially novel; a given stimulus is
used on only a single trial (“trial unique”). A novel stimulus is presented as the sample, and, after a delay last-
FAMILIAR
STIMULI
ing up to several minutes, it is presented again as a (familiar) test stimulus, together with a new novel stimulus. Although the animal must indicate which of the two test stimuli matches the sample, as in the recency design, it may also
base its choice on the relative novelty or familiarity of the
two test stimuli.
Several neurophysiological
studies have investigated recency effects in IT cortex with the use of the DMS paradigm
with familiar stimuli and with delays of up to several seconds. These studies have reported that many IT neurons
respond differently to a test stimulus depending on whether
or not it matches the sample (Baylis and Rolls 1987; Eskandar et al. 1992; Gross et al. 1979; Mikami and Kubota
1980; Miller et al. 199 1b; Riches et al. 199 1; Vogels and
Orban 1990), and this difference is maintained even when
several other stimuli intervene in the retention interval
( Miller et al. 199 1b, 1993). Because the responses of many
cells to a current stimulus are suppressed according to its
similarity to the sample held in memory, we have suggested
that IT cells function as “adaptive mnemonic filters.”
Whether IT neurons might also play a role in recognition
memory has been less clear. Studies of IT neurons in anesthetized or passively fixating monkeys report that the responses of some cells “habituate” to repeated presentations
of a stimulus (Gross et al. 1972; Pollen et al. 1984; Richmond et al. 1983) with effects lasting no more than 12 s
( Miller et al. 199 la). Recognition memory, however, requires a neural mechanism that is both stimulus specific
and longer lasting than 12 s. Baylis and Rolls ( 1987) studied neurons in the lateral portion of IT cortex in animals
performing a serial recognition memory task. They found
that the responses to novel stimuli decreased with repetition, but this effect was eliminated if more than two other
stimuli intervened in the retention interval. By contrast,
Riches et al. ( 199 1) recorded from a few cells in the anterior
ventral portion of IT cortex and found changes in response
to novel stimuli that were maintained over at least several
intervening stimuli. Finally, Rolls et al. ( 1989) reported
that the responses of some face-selective cells in IT cortex
systematically increase for some faces and decrease for
others as initially novel faces become familiar. However, it
is not known if this phenomenon is specific to faces.
To explore further the role of IT cortex in visual recognition memory, we studied IT neurons in fixating monkeys
performing a version of the DMS task. There were two critical differences in the methods of the present study compared with those used in our study of recency memory in IT
cortex ( Miller et al. 199 1 b, 1993). First, whereas the stimuli
in the recency study were all familiar to the animal, most of
the stimuli used for each cell in the present study were initially novel. Second, whereas the focus in the recency study
was on within-trial modulation
of responses to matching
and nonmatching stimuli, the focus of the present study
was on across-trial effects, namely the responses to the initially novel stimuli as they gradually became familiar to the
animal during the recording session. Because we did not
require the animal to distinguish novel from familiar stimuli, any effects of increasing familiarity on the cells’ responses would be due to incidental memories that accrued
during the session. In a previous preliminary report we described how IT neuronal responses decline with stimulus
AND
IT CORTEX
1919
familiarity (Miller et al. 199 1b). In the present study we
describe several manipulations
that help define how this
familiarity mechanism operates in IT cortex.
METHODS
Subjects
Two rhesus monkeys weighing 8-9 kg were used. These were
the same animals used in the study of recency memory in IT
cortex (Miller et al. 1993) and the surgical, recording, and histological methods were also the same.
Behavioral
task
The task was a modified version of the DMS task used to study
recency memory ( Miller et al. 1993). The monkey began each
trial by grasping a metal bar and fixating a small (0.2’) fixation
target, which remained on throughout
the trial. Trials were
aborted if the monkey’s gaze deviated from the fixation target
> l-l So at any time during the trial (fixation typically varied
much less than 1O). A sample stimulus was presented 300 ms after
the animal achieved fixation. The sample was followed by two to
five test stimuli. When one of the test stimuli matched the sample,
the monkey was required to release the bar within 900 ms of stimulus onset to receive an orange juice reward. Each stimulus was
presented for 500 ms, and there was a 700-ms delay between the
disappearance of one stimulus and onset of the next. The match
stimulus was the last stimulus presented on each trial, and this
stimulus was extinguished
as soon as the monkey made its response, typically -350 ms after stimulus onset. The time between
trials was - 1-3 s, depending on how long the animal took to
drink the juice and initiate the next trial. We searched for cells
while the animal performed the DMS task, with the use of a set of
familiar sample and test stimuli. Once a cell was isolated, we
switched to the novel set and initiated the data acquisition.
A new set of 20 initially novel stimuli were used as the samples
and matching test stimuli for each cell, and a fixed set of four
highly familiar stimuli were used as the nonmatching
ones (the
same nonmatching
items were also used in the task used to search
for cells). Thus the initiallv novel stimuli appeared only at the
beginning (sample) and end ( matching test) of each trial, never as
an intervening
nonmatching
stimulus. This design was chosen
because it greatly simplified the analysis of repetition and temporal order effects for the initially novel stimuli. Although the use
of familiar nonmatching
stimuli allowed the animal to possibly
use relative novelty within a trial as a cue to whether a test stimulus was a match or nonmatch, our major interest was in the changing response to the sample stimuli rather than match-nonmatch
effects. The main purpose of the behavioral task was to ensure that
the animal attended to the stimuli and held them in memory, at
least temporarily.
The stimuli were complex, multicolored
pictures digitized from
magazine photographs and presented on a computer display. The
images were of faces, bodies, natural objects, and complex patterns that did not appear in the laboratory setting. We did not
attempt to find “optimal”
stimuli for any of the cells. Rather, we
randomly selected a wide variety of complex pictures for each set,
which previous studies indicated would elicit a full range of responses from any given cell (Desimone et al. 1984: Gross et al.
1972: Tanaka et al. 199 1). The stimuli ranged between 1 and 3’
on a side and were presented at the center of gaze. Most of the
stimuli were used for only one cell, but a few were reused after a
minimum of 3 mo.
The initially novel sample stimuli were presented in a particular
19x
L. LI,
E. K. MILLER,
AND
R. DESIMONE
Trial #I
Trial #2
FIG. 1. Schematic
representation
of trial sequence. Initially
novel stimuli were used as samples and matching
test stimuli.
Nonmatching
stimuli are represented
by dots. A given sample
stimulus
is repeated
after 3 or 35 intervening
trials, in alternation.
The apple, for example, appears on trials 1, 5, 4 1, 45, 8 1, etc.
Trial #3
Trial #4
Trial #5
4
Cl
I
I
I
Trial #41
l
4
cl
fixed order, allowing comparison of the responses to strmun repeated after either a small or large number of intervening trials, as
follows. The 20 stimuli were randomly divided into 5 blocks of 4
stimuli each. The first block of four stimuli was used as the samples on the first four trials of the session, and then this same block
of stimuli was repeated in the same order for the next four trials of
the session. We then switched to the second block of stimuli,
which was also used for two sets of four trials, and so on. After all
20 stimuli appeared in 2 blocks (i.e., on 2 trials) each, we then
repeated the whole series several times, for a total of 1 l-20 trials
per stimulus. Thus the number of intervening trials between repetitions of a given sample alternated between 3 and 35, and each
sample stimulus appeared on an intervening trial for every other
sample stimulus. The sequence of trials is illustrated in Fig. 1.
Across all types of trials, the average number of stimuli presented per trial was 4.3. Thus, in the blocks with 4 trials, a given
stimulus was repeated, on the average, after 13 intervening stimuli
( 3 intervening trials), which took -25 s. In blocks with 35 intervening trials, a given stimulus was repeated, on the average, after
152 intervening stimuli, which took -5 min. A complete session
with a single cell typically lasted - 1 h.
Dutu unulwis
.
Responses to stimuli were appraised by the use of analysis of
variance (ANOVA),
f tests, and linear regression, evaluated at the
P < 0.05 level of significance. Responses to match stimuli were
always calculated over a 200-ms time interval beginning 75 ms
after stimulus onset. The beginning of this time interval was chosen to coincide with the typical minimum response latencies of IT
neurons, and the end was chosen to occur well before the animal’s
behavioral response (mean reaction time was 359 ms). To facilitate comparison with the match responses, responses to the samples were also calculated over this 200-ms interval, except for the
linear regression computed on sample responses over the session.
For this latter test, we utilized the full neuronal response for the
sample, which was measured over a 425-ms interval beginning 75
ms after stimulus onset. To compare response histograms averaged from multiple cells, the individual spike trains for each cell
were first “smoothed”
by convolution
with a Gaussian with a
standard deviation of 10 ms (Richmond
and Optican 1987).
RESULTS
Anatomic
location ofpenetrutions
.
As shown in Fig. 2, all of the recording sites were located
on the inferior temporal gyrus, just lateral to the rhinal sulcus and medial to the anterior medial temporal sulcus.
General propert ies
A total of 72 visually responsive neurons were recorded.
All neurons gave excitatory responses. To determine
whether the response to a sample stimulus was significantly
different from the neuron’s spontaneous activity, we compared the firing rate during the presentation of each sample
stimulus with the prestimulus firing rate with the use of a
paired t test. Of the 1,440 stimuli used as a sample (72
neurons X 20 stimuli),
967 (64%) elicited a visual response. To determine whether a given neuron was stimulus
selective, we performed an ANOVA on the responses to the
20 sample stimuli. The ANOVA was significant (P < 0.05)
for 89% (64/72) of the neurons.
Efict
oj’familiarity I on the responses to the sample stimuli
The responses of many IT neurons changed as the initially novel stimuli were repeatedly presented throughout
the session. Figure 3 shows, for 1 cell, response histograms
for the 1st through 6th presentations of all 20 sample stimuli averaged together. The responses to the stimuli were
strongest when they were novel and became progressively
weaker as they were repeated throughout the session, i.e., as
they became familiar. We will refer to this change in response with stimulus repetition as the “familiarity
effect.”
For each cell, we determined whether there was a progressive change across the session in response to the samples by
computing a linear regression on the response to the first 12
presentations of the 20 sample stimuli. Responses to all
samples were used in the regression, including samples that
may not have elicited any significant response (relative to
FAMILIAR
STIMULI
FIG. 2. Location of recording zones (a) in both monkeys. AMT, anterior middle temporal sulcus; CA, calcarine sulcus; IO, inferior occipital
sulcus; OT, occipitotemporal sulcus; ORB, orbital sulcus; RH, rhinal sulcus; ST, superior temporal sulcus.
1921
AND IT CORTEX
(according to the linear regression described above). The
trial number on the abscissa is the absolute trial number in
the session, with the numbering beginning with trial 0.
Thus a stimulus appeared as a sample for the first time on
trial number 0, for the second time on trial number 4, for
the third time on trial number 40, for the fourth time on
trial 44, and so on. Across the population of cells, there is a
waning of response that reaches a plateau after six to eight
sample presentations (because each stimulus also appeared
as a match at the end of a trial, this corresponds to 12-l 6
stimulus presentations). For comparison, the figure also
indicates the neurons’ average baseline, or spontaneous, activity for the same trials. The response to the first presentation of a stimulus was - 20 spikes/s above the baseline rate,
whereas the plateau was only - 8 spikes/s above the baseline, or -40% of the initial response. This reduction to 40%
of the initial response is probably an underestimate of the
size of the familiarity effect, because the average is based on
the response to all stimuli, some of which elicited little or no
response throughout the session. The baseline rate did not
show any significant change over the session.
The staircase appearance of the line relating response
magnitude to trial number in Fig. 4A is due to the two
different numbers of trials intervening between presentations of a given sample stimulus. The number of intervening trials between repetitions of a given sample alternated
between 3 and 35 throughout the session (see METHODS).
Figure 4A shows that the decrement in response was greater
when only 3 trials intervened than when 35 trials intervened between presentations of the same sample stimulus.
The response to a specific novel stimulus “A ,” for example,
declined more when the animal had seen A 4 trials ago (3
intervening trials) than 36 trials ago. Closely spaced repetitions are apparently more effective at “stamping in” the
memory trace. However, even though there was no further
decrement after 35 intervening trials, the response did not
J139201
1201
-
the spontaneous firing rate) during the session. On the basis
of this regression, about one-third of the neurons (25 /72)
exhibited a significant decrease in response across the session, and 18% ( 13/72) exhibited a significant increase in
response. We will first describe some of the properties of the
cells showing a decrement in response with familiarity and
then consider the cells that appeared to show an increase.
Decrement of responsewith Jkniliarity
Of the 25 cells with a significant decrement of response
with familiarity, 24 of them (96%) were stimulus selective
according to the ANOVA described above. These cells did
not simply respond equally to all novel stimuli and thus
were not “novelty detectors” in that sense. An optimal response seemed to require a stimulus that was both novel
and of the appropriate color, shape, and so on.
Figure 4A shows the responses to the first 11 presentations (the minimum number of presentations per cell) of all
20 sample stimuli, averaged from all the responses of the 25
IT neurons that showed a significant overall decrease in
response with repeated presentation of the sample stimuli
FIG. 3. Example of responses of an individual inferior temporal (IT)
neuron to a set of 20 initially novel sample stimuli. Responses to all 20
stimuli have been averaged together into single composite histograms. The
6 histograms show the average response to the 1st 6 presentations of all
stimuli. Horizontal line under the histograms indicates when the stimuli
were on. The vertical scale indicates firing rate in spikes/s, and the binwidth is 10 ms.
1922
L. LI,
E. K. MILLER,
Sample
A
u
c
8
a,
cn
30
‘t
Baseline
Match
. ...* AP
25
l
.
.
.
20
tii
Q
4
g
Y
-a
co
l
15
;...*
l
***
A
l
4
10
5
B
I
I
0
40
L
80
I
120
I
4
160
200
Nonmatch
-0 * 0,
30
0
40
Number
80
120
of trials measured
160
200
from first trial
FIG. 4.
A: average responses of 25 neurons whose responses declined
significantly
with increasing
stimulus
familiarity
over the session. Solid
line indicates
responses to the sample stimuli,
dotted line indicates
responses to the match stimuli, and dashed line gives baseline (prestimulus)
firing rate. Trial number is measured from 1st trial of a given stimulus.
B:
average responses of the same cells to a set of 4 familiar
nonmatching
stimuli. Baseline rate is also shown, for comparison.
recover to its initial value. Thus the cells “remembered”
a
stimulus even after 152 intervening stimulus presentations
(35 intervening trials with an average of 4.3 stimuli per
trial), or a retention interval of -5 min.
The larger decrement in response after 3 trials than after
35 trials establishes that the response decrement is stimulus
specific and cannot be due to neuronal “fatigue” or other
nonspecific changes that might have occurred during the
session. If the neurons were simply becoming fatigued (or if
the decrement were not linked to specific stimuli), there
should have been a larger response decrement after 35 intervening trials than after 3 intervening trials, because more
time elapsed, more stimuli appeared, and more responses
occurred with the larger number of intervening trials.
To determine how many cells showed a significant drop
in response after just three intervening trials, we performed
a paired t test on the responses to each sample stimulus
before and after each block of three intervening trials. For
this comparison, any overall changes that occur over the
entire recording session are discounted. On the basis of the t
AND
R. DESIMONE
test, 13 ( 18%) of the 72 total cells showed a difference in
response to the same stimuli repeated after 3 intervening
trials. All of these cells were in the group of 25 that showed a
significant trend in response over the session based on the
linear regression, and all of the response differences were
decrements. Thus the only consistent “short-term” change
in response to repetition in IT cortex was a decrement. The
fact that fewer cells showed a significant short-term change
in response than a significant overall trend is presumably
due to some cells showing consistent, but not statistically
significant, short-term decrements in response that accumulated over the course of a session, resulting in a significant trend.
If the decrement in IT neuronal responses was due to
increasing familiarity, we would expect that the responses
to the four familiar nonwlatching stimuli presented on most
trials would show little decrement. The same nonmatching
stimuli were used for several months over the course of the
experiment, and they had also been seen by the animal at
the start of each recording session when we searched for
cells. Figure 4B shows the average responses to the nonmatching stimuli, for the same 25 neurons that showed a
decrement in response to the samples. The population response to the familiar nonmatching stimuli declined only
slightly over the session, and none of the cells showed a
significant trend.
Because the nonmatching stimuli were already familiar
to the animal at the start of each recording session, we
might also expect the response to these stimuli to be
smaller, on the average, than those to the first few presentations of the initially novel stimuli used as samples and
matches. This comparison is complicated by the fact that
IT responses to familiar stimuli are typically largest to samples, followed by nonmatching and then matching stimuli
(Miller et al. 199 1b, 1993). However, it can be seen by
comparing A and B of Fig. 4 that, on the initial three presentations of the novel stimulus set (through trial 40), responses of the population of cells to the familiar nonmatching items were not only smaller than those to the novel
sample but also smaller than to the novel matching test
stimuli. This reversal of the order of responses to matching
and nonmatching stimuli that had previously been seen in
studies with familiar stimuli is likely due to the novelty of
the matching stimuli in the present study. After the first
three presentations of the stimulus set, when the initially
novel stimuli were now somewhat familiar, responses to the
matching stimuli fell below those to the nonmatching,
which is the typical order of responses to familiar stimuli.
Stimulus
selectivitv w ofthc
.
.familiarity
efict
..
As described above, the fact that the familiarity effect was
larger when 3 trials intervened between repetitions of a sample than when 35 trials intervened indicated that the familiarity effect was stimulus specific. As an additional test of
stimulus specificity, we tested 10 cells with a 2nd set of
novel stimuli after completing a session with the 1st set.
Figure 5 shows an example of a single IT neuron’s responses
to presentation of 1 set of 20 novel stimuli and the neuron’s
subsequent responses when a new set of stimuli was intro-
FAMILIAR
STIMULI
24
20
18
16
J264101
\
I
I
I
240
280
320
from First Trial
1
360
;
14'
IT CORTEX
1923
Cells showing an increase in response with .familiarity
22
-z
8
g
&
a
$
Y
-a
c/)
AND
'
0
I
40
I
80
Number
I
I
I
120
160
200
of Trials Measured
FIG. 5.
Average responses of an individual
IT neuron to 2 sets of initially novel stimuli. After the neuron’s response to the 1st set had significantly declined (-),
a new set of stimuli was introduced.
The response
rebounded
(- - -) and then declined again as the new set became familiar.
duced. The responses of this neuron declined as the first set
of stimuli became familiar. When a new set of novel stimuli
was then substituted for the first, the neuron’s response recovered to a high rate, followed by a progressive decrement
in response as the new stimulus set became familiar. Thus
the response decrement was tied to the familiarity of specific stimuli, and it did not generalize to other, unseen, stimuli. This recovery of response to new stimuli was seen for all
10 cells tested with a 2nd set.
Relationship
bet weenjbmiliarit
and matching eflticts
wv efjkt
..
In our previous study of IT cortex in monkeys performing a DMS task, we found that responses to test stimuli that
matched the sample held in memory on a given trial were
suppressed in comparison with the responses to samples
and nonmatching stimuli (Miller et al. 199 1b, 1993). To
determine whether the neurons that showed the (acrosstrial) familiarity effect also showed the (within-trial)
suppression of response to matching stimuli (“matching effect”), we examined the average responses to the matching
stimuli at the end of each trial. As shown in Fig. 4A, the
average response to matching stimuli was weaker than the
average response to sample stimuli throughout the session.
Furthermore, the responses to the matching stimuli declined across the session, in parallel with the decline in responses to the samples. The within-trial matching effect appeared to add to the across-trial familiarity effect. Thus an
IT neuron appears to be capable of coding both types of
mnemonic information: the matching status of a stimulus
within a trial and the relative familiarity of the stimulus that
builds up across trials.
A relationship between familiarity and recency is further
suggested by the fact that the 25 cells showing a decrement
in response with familiarity had a larger matching effect
than the remainder of the population of cells. The response
to the match stimulus was only 67% of the sample response
in these 25 cells, compared with 94% ofthe sample response
for the remainder of the population. This suggests that information about both familiarity
and recency may be
carried mostly by the same subgroup of IT cells.
The stimulus specificity ofthe familiarity effect, the stability of the baseline activity across the recording session, and
recovery of response with a new set of stimuli (for the 10
neurons so tested) all indicate that the response decrement
over the session was not caused by nonspecific changes in
the cortex, such as damage or irritation to the recorded cells
or surrounding neuropil. By contrast, the 13 cells showing a
significant increase in response over the recording session
could well have been due to such nonspecific factors. Figure
6A shows the average response to the 1st 11 presentations
of all 20 samples for these 13 cells, as well as their baseline
activity. The responses and baseline activity of the 25 cells
with a response decrement are also shown, for comparison.
Unlike the cells showing a decrement, whose baseline activity was constant, the cells whose responses increased over
the session showed parallel increases in their baseline firing
rates, suggesting some change in overall excitability, possibly due to damage or irritation. This conclusion is supported by the shape of the curve showing the average response over the session for these cells. The curve has a negative staircase appearance, superimposed on a larger, more
gradual, increase. That is, with three intervening trials between presentations of a given sample, repetition actually
appeared to cause a decrement in response of these cells,
A
30 '
z
25
8
CD
a
ii5
a
20
51
B
’
0
I
I
I
I
I
40
80
120
160
200
4.00
“negative
cells”
“positive
cells”
I
$I 3.50
.Q)
ati 3.00
-2
E
0 2.50
5f
E
2.00
I
’
I
I
I
I
0
40
80
120
160
Number of trials measured
FIG. 6.
200
from first trial
A : average response of 13 IT neurons whose response increased
significantly
over the course of the recording
session, or positive cells (top
- - -) and, for comparison,
the average response of the 25 cells whose
response significantly
decreased, or negative cells (--).
The prestimulus
baseline activity for the 2 groups is shown by the 2 bottom lines. B: average
response of the same cells, but with the response divided by the baseline
firing rate.
L. LI, E. K. MILLER,
whereas there was a net increase in response over the longer
number of intervening trials.
To control for nonspecific changes in excitability, as
measured by changes in the baseline activity, we divided the
response to each stimulus by the average baseline firing rate
on each trial for both the “positive” and “negative” cells.
Figure 6B shows the stimulus responses, which are now
expressed as multiples of the baseline firing rates. Although
the “negative cells” still show a decrease in response over
the session, the “positive cells” no longer show an increase
but actually show a slight decrease. If the responses of these
few cells with increments are indeed due to nonspecific
changes, then the only change in response with familiarity
in IT cortex may be a decrement.
This conclusion is further supported by the results shown
in Fig. 7, which shows the average responses of the entire
population of recorded cells, excluding the 25 cells with a
significant decrement in response. There is no tendency for
the remaining neurons to show any increase in response
over the session. Interestingly, the average response of the
cells unaffected by familiarity was 15.24 spikes/s, which
was only about one-half the response of the 25 “decrement”
cells to the initial presentation of a novel stimulus. In fact,
the average response of the unaffected population
was
about equal to the “plateau” response of the decrement
cells, after the initially novel stimuli had become familiar
for that group.
o/‘fbmiliurit
..
Iv on responses to individual stimuli
Even though there was little or no evidence for a subpopulation of IT cells whose average response to all stimuli increased with familiarity, it is still possible that the responses
of some cells increased for just a few of the stimuli tested,
but did not change, or even decreased, for other stimuli.
Such cells might have been missed in the regression analysis, which was based on the average response across all
stimuli.
To test whether the change in response over the session
was different for different stimuli, we computed a multiple
linear regression for each cell, with number of repetitions as
a linear factor and the particular stimulus tested as a classifiEJixts
30 25 73
6 203
l
tl
a 15YfG
g
lo-
nm---....--------a D---...-.e ------a
w----------
5-
0
’
0
Baseline
----~~---...“-------.=----
I
I
I
I
40
80
120
160
----___ --__l
1
200
Number of trials measured
from the first trial
Average response of the entire sample of IT neurons, excluding
the cells that showed a significant
decrease in response with familiarity.
FIG.
7.
AND
R. DESIMONE
100 1
.
2nd
--a--.. ..6th
.. .. . .. .
10th
e-0
-. -” -..-.. -.-..-.-..-..-..-
Baseline
..-..-..-..-.UII
--..-.-.-
40
80
Number
of trials measured
from first trial
FIG. 8.
Responses of an IT neuron to 4 stimuli. The 20 novel stimuli
tested were ran ked according t their in it ial respo nse, and the lines indicate
the responses t o the lst, hd,
th, and 0th best of the 20.
0
cation factor. We then asked whether there was a significant
interaction between the stimulus (stimulus factor) and the
rate of change of response (repetition factor). Because stimuli that elicited no response could not show a response
change over the session, we limited this test to the 10 “best”
stimuli for each cell. These stimuli were determined by
rank ordering the stimuli on the basis of the average response to the first two presentations.
The interaction was significant for eight cells (out of the
72 total). That is, the rate of change in response across the
session for these cells was different for different stimuli.
One of the eight showed a significant overall increase in
response with familiarity,
which was correlated with a
change in the neuron’s baseline firing rate. We did not include this cell in further analyses. For the remaining seven
cells, we computed a simple linear regression of the response across repetitions, for each stimulus separately. Of
the 70 stimuli used (7 cells X 10 stimuli), the regression was
significant for 24, and all of these showed a decrease in
response across the session. No individual stimulus showed
a significant increase in response over the session.
Inspection of the responses for the seven cells with significant interactions between the stimulus factor and the rate of
response change indicated that these interactions were
caused by differences in the initial responses to the stimuli.
Specifically, the stimuli that elicited a stronger initial response showed a greater response decrement over the session than did less effective stimuli. The significant interaction in the regression was apparently due to a “floor” effect
for the stimuli that initially elicited poor responses. An example of 1 cell exhibiting an interaction is shown in Fig. 8,
which shows the responses of the cell to the 1st 2nd, 6th,
and 10th best stimuli. The initial responses to all four of the
stimuli were well above the neuron’s baseline firing rate.
However, the responses to the 1st and 2nd best stimuli
showed sharp declines with repetition, whereas the responses to the 6th and 10th best stimuli showed little or no
change. Because the decrement in response was largest for
the stimuli that initially gave the best response, the responses of the cell showed greater selectivity among the
FAMILIAR
STIMULI
1st
-------e
3rd
10
+““““““““““”
0
-30
30
60
120
90
180
150
240
210
300
270
Time from stimulus
360
330
onset
420
390
480
450
540
510
570
(msec)
FIG. 9. Time course of the response to novel stimuli as they became
familiar. Curves show the responses to the lst-6th presentation of all 20
initially novel stimuli, for the 25 cells that showed a significant decrement
in response with familiarity. The stimulus was turned on at time 0 and
turned off 500 ms later.
stimuli when they were still relative1 y novel than when they
were familiar, i.e., after fou r to five presentations.
Time course of. the .fumiliarity
e$Gct
To determine how soon after the onset of the visual response IT cells distinguished among novel and familiar
stimuli, we computed average response histograms for each
of the 1st 6 presentations of all 20 sample stimuli for the 25
cells with a significant familiarity effect. Fig. 9 shows the
response histograms for all 25 cells, averaged together. The
responses to the first and second presentations of the same
sample stimulus (with 3 intervening trials) have similar
magnitudes and time courses until relatively late into the
respon se, at which point the response to the second presentation appears to become suppressed. By contrast, the responses to the third through sixth stimulus presentation diverge from that of the first almost at the beginning of the
visual response. Thus, by the third sample presentation, IT
cells appear to distinguish between novel and familiar stimuli almost immediately. It should be noted that responses to
the second versus third presentations as well as to the fourth
versus fifth presentations show little or no divergence. This
is due to the fact that 35 trials intervene between the 2nd
and 3rd presentations and the 4th and 5th presentations,
and IT cells show little or no additional decrement in response to a stimulus after 35 intervening trials. By contrast,
responses to the third versus fourth and fifth versus sixth
presentations do diverge, with three intervening trials.
To quantify these differences in the time course of the
response, bin by bin paired t tests were computed for several
pairs of presentations ( 1st vs. 2nd, 3rd vs. 4th, etc.). Response differences were taken to begin in the earliest bin in
which the paired t test reached significance (P < 0.05 ). The
latency of the visually evoked response was determined by a
paired t test between the average response across the six
presentations and the average baseline activity. For each
presentation the response began -70 ms after stimulus onset, that is, the paired t test first reached significance in the
bin corresponding to 70-80 ms after stimulus onset. The
AND
IT CORTEX
1925
responses to the first and second presentations of the stimuli became significantly different 170- 180 ms after stimulus onset, or 1OO- 110 ms after response onset. By contrast,
the difference in response between the first and fourth stimulus presentations became significant 80-90 ms after stimulus onset, which was only 10 ms after the start of the response. Thus, as the stimuli become increasi
familiar,
the latency of the familiarity effect is reduced. The difference in response between the third and fourth and between
the fifth and sixth presentations occurs 1OO- 110 and 11O120 ms, respectively, after stimulus onset, or 30-50 ms after
the beginning of the response.
We also computed average response histograms for the
matching stimuli at the end of each trial, to compare with
the time course of the familiarity effect for the samples.
Figure 10 shows the average responses to the 1st and 2nd
presentations of all 20 sample and matching stimuli, for the
25 cells showing a response decrement with familiarity. Although the response difference between the first and second
sample presentations occurs relatively late, the difference
between the sample and match responses occurs immediately from the onset of the visually evoked response. The
immediate suppression of the match response is compatible
with the results of our previous study of IT cells (Miller et
al. 1993)‘) in which we found immediate suppression of the
match response compared with the nonmatch.
DISCUSSION
The responses of many IT neurons are determined by
both the sensory features and familiarity of stimuli. About
one-third of the neurons in our sample gave their strongest
responses to specific novel stimuli and progressively weaker
responses as the stimuli became more familiar over the
course of an hour-long session. The cells are not novelty
detectors, in the sense of cells that respond to any novel
stimulus (Wilson and Rolls 1990). Rather, sensitivity to
novelty seems to be accomplished in parallel with sensitivity to conventional visual features such as color, shape, etc.
In a similar vein, we previously found that the responses of
IT cells to otherwise effective test stimuli are suppressed
according to their similarity to the sample stimulus held in
35
1st presentation
30
2nd
presentation
. . . . . .. .. . . . . . .
z
8 25
%
iii
a 20
iE
Y
'a 15
cn
10
”
0
-30
60
30
120
90
180
150
240
210
300
270
Time from stimulus
360
330
onset
420
390
480
450
540
510
570
(msec)
FIG. 10.
Average response histograms to the 1st and 2nd presentations
of the initially novel sample and matching test stimuli for the 25 cells that
showed a significant decrement in response with familiarity. See Fig. 9 for
conventions.
1926
L. LI, E. K. MILLER,
memory on a given trial of a DMS task (Miller et al. 199 lb,
1993). Together with the results of the present study, the
findings indicate that current stimuli are compared with
both short- and long-term memory traces in IT. On the
basis of this property, we have previously described IT cells
as adaptive mnemonic filters.
We reached this conclusion about effects of stimulus familiarity in an earlier paper (Miller et al. 199 1b), but the
analyses of the present study allow us to expand on the
previous one. We now have further evidence that the familiarity effect is stimulus specific; that it is largest for stimuli
that initially cause the greatest response; that it is rarely, if
ever, associated with an increase in response; that it affects
responses to both sample and matching test stimuli in parallel; that it tends to occur in the same cells that show the
largest repetition effects within a trial; that it has little effect
on responses to stimuli that are already familiar; that it
begins earlier in time as stimuli become increasingly familiar; and that it ultimately begins within 10 ms of response
onset.
Relationship
to prior studies
Although it has long been established that IT cortex is
important for visual memory, recent studies suggest that
the most anterior-ventral portion of IT plays the most important role. This region includes “perirhinal”
cortex,
which is comprised of area 35 in the lateral bank of the
rhinal sulcus and area 36, located in the cortex between the
rhinal and anterior middle temporal sulci and extending
into the temporal pole (Amaral et al. 1987; Suzuki et al.
1993). Lesions that include perirhinal cortex have a devastating effect on performance of DMS tasks (Gaffan and
Murray 1992; Horel et al. 1987; Meunier et al. 1990;
Murray 1992; Murray et al. 1989; Suzuki et al. 199 1, 1993;
Zola-Morgan et al. 1989, 1993). Horel et al. ( 1987) directly
compared the effects of cooling the anterior-ventral portion
of IT cortex, including area 36 in the perirhinal region, with
cooling lateral area TE and found a greater impairment on
a DMS task from the ventral site. Although we were not
able to localize recording sites precisely enough to distinguish ones in area 36 from those in TE, our recording region appears to span both regions. In future studies it will be
important to compare cell properties in this region with
those in adjacent areas such as lateral TE, area 35, TF/TH,
and the hippocampus.
Although the role of IT neurons in recency memory has
been investigated in a number of studies, there have been
relatively few studies of the role of IT neurons in visual
recognition. Baylis and Rolls ( 1987) tested the responses of
IT neurons to novel stimuli in a serial recognition task and
found that responses to novel stimuli typically declined
when they were repeated. However, this effect was eliminated for most cells if even a single stimulus appeared in the
retention interval and was eliminated for all cells with two
intervening stimuli. They concluded that IT cortex “would
be useful for short-term visual memory . . . but would not
be useful in recency memory tasks in which one or more
stimuli intervenes between the first and second presentations of a stimulus.” Miller et al. ( 199 1a) reported that the
responses of IT neurons in both anesthetized and passively
AND
R. DESIMONE
fixating (awake) monkeys declined with stimulus repetition, but only if the stimuli were repeated at least once every
12 s. The failure to find long-lasting repetition effects in
either of these two studies could be due to differences in
either the task used or in the recordings sites. The latter
seems more likely, because neither of these two studies included cells in the anterior-ventral portion of IT cortex included in the present study.
Riches et al. ( 199 1) did record in this anterior-ventral
region and found that the response decrement to repeated
stimulation survived intervening stimuli. However, the effect of intervening stimuli was tested in only a few cells, and
the time course of the effect was not clear. Our data indicate
that these familiarity effects can survive retention intervals
lasting at least many minutes, filled with > 100 stimulus
presentations. Riches et al. ( 199 1) also reported that some
IT neurons gave stronger responses to a set of novel stimuli
than to a set of highly familiar ones. Although this response
difference might have been caused by featural differences
between the stimuli in the novel and familiar sets, the fact
that we found larger responses to the novel sample and
match stimuli than to the highly familiar nonmatching stimuli suggests that the difference was in fact due to long-term
memories of the familiar stimuli.
Finally, Rolls et al. ( 1989) recorded the responses of faceselective cells in the superior temporal sulcus and lateral
portions of IT cortex to initially novel faces that were repeatedly presented. The responses of some cells gradually
changed with repetition, even when repetitions of a stimulus were separated by several intervening items. This finding is consistent with the present results, except that neurons in anterior-ventral
IT are not especially selective for
faces. We found familiarity effects for a wide variety of stimuli, suggesting that this anterior-ventral
region is a more
“general purpose” recognition zone. Rolls et al. also found
that the responses of face-selective cells decreased with repetition of some faces but increased with others, a result that
we did not find for other objects in anterior-ventral IT cortex. This difference again could be due either to a difference
in recording sites or to differences in the experimental design. Unlike in the present study, the monkeys in the Roll et
al. study were not required to remember the faces for even a
short period.
Adaptive memory cells and multiple
memory systems
There have been many different types of memory defined
at the behavioral level, which can generally be placed in two
broad classes. One major class is explicit, or declarative,
memory, which is the memory of specific facts and events.
Another class is implicit,
or nondeclarative,
memory,
which includes stimulus-response habit learning, skill learning, perceptual learning, and various types of nonassociative memory such as priming, habituation, sensitization,
and so on. Behavioral studies suggest that IT cortex plays a
role in both classes of memory. Lesions of IT cortex, particularly the anterior-ventral
portion, impair performance of
DMS tasks, which are thought to tap explicit memory (Gaffan and Murray 1992; Horel et al. 1987; Meunier et al.
1990; Murray 1992; Murray et al. 1989; Suzuki et al. 199 1,
1993; Zola-Morgan et al. 1989, 1993). Likewise, IT lesions
FAMILIAR
STIMULI
impair visual discrimination
learning, which is thought to
involve implicit memory, at least in part (Mishkin
1982;
Phillips et al. 1988).
As we have argued previously, adaptive memory filtering
in IT might contribute to any or all of the memory systems
defined in behavioral studies, depending on how the memory system is implemented
neurally (Miller et al. 1993).
Recently, we considered the possible contributions of IT
neurons to performance of a recency memory task. We proposed that neurons in a hypothetical “decision network”
downstream from IT receives opposing inputs from two
different sets of IT neurons ( Miller et al. 1993 ) . One of the
input sets (sensory cells) codes the sensory properties of the
stimulus, with no influence of memory, and the other
(adaptive cells) gives reduced responses to stimuli depending on how similar they are to stored memory traces. A
difference in mean response between the two sets of cells
signals that the current stimulus is a “match.” The present
results suggest that the same sort of decision network could
support judgements of familiarity. A difference in response
between the adaptive cells that show the familiarity effect
and cells that do not could signal that the current stimulus
is a familiar one. When both novel and familiar stimuli are
used in a matching to sample task, the recency and novelty
mechanisms may interact or even interfere. We found that
responses to novel matching stimuli were larger than to
familiar nonmatching stimuli early in the session, and a
reversal of this relationship later in the session when the
matching stimuli were more familiar. It is possible that the
animal solved the task at the beginning of the session on the
basis of relative novelty or familiarity (e.g., release the response bar to any novel test stimulus) and later switched to
a recency strategy (e.g., release the bar to any stimulus repeated within a trial).
Because fewer IT cells respond strongly to a stimulus
after it has become familiar, familiarity may cause a “focusing” of activity across IT cell populations. Cells that poorly
represent the features of a given familiar stimulus may be
winnowed out of the activated population through the
adaptive memory mechanism, in much the same way that
an excess of cells and connections are lost during development. Ironically, then, it may be the cells that do not show
response decrements with familiarity that actually contribute to the long-term memory of a familiar stimulus. If these
remaining activated cells have the appropriate association
with cells coding contextual information
(the circumstances in which the stimuli were seen, for example), they
might mediate, in part, an “explicit” memory of the stimulus.
On the other hand, the adaptive memory mechanism
might mediate priming phenomena, which reflects implicit
memory. In a typical priming task for visual patterns, a
subject is first shown a list of drawings, without any instruction to remember them. Later, they are given a picture recognition task, and their performance is usually faster or
better for the stimuli that they had seen before, even if they
have no conscious memory of having seen them (Schacter
et al. 1990, 199 1; Squire et al. 1992). It is commonly believed that priming is due to a tendency of neuronal populations to be more easily activated if they have been activated
previously. For individual cells, we have shown that this is
1927
AND IT CORTEX
Sensory
Input
K
Adaptive
Visual Feature
Analyzers
At-tent i!;
Control
Orienting
Systems
Memory
Filters
7
Interaction between systems for memory and attention. In
FIG. 1 1.
this scheme, a novel stimulus will activate adaptive memory filter cells in
IT cortex, which in turn will drive attentional and orienting systems. This
will lead to increased attention and contact with the new stimulus, causing
adaptation of synaptic weights in IT cortex, reducing the activation of the
cells. When the novelty of the new stimulus and the activation of IT cortex
is sufficiently diminished, the system will be ready to process other competing stimuli.
not the case, at least in IT cortex. However, as we suggest
above, it is the elimination
of certain cells from the activated population that may be important in forming the
underlying neuronal representation of a stimulus. Repetition may speed the construction of this critical population,
resulting in faster and better recognition when a stimulus is
repeated. Results from a recent positron emission tomography (PET) study of priming are consistent with the idea that
priming is associated with a reduction of activity in the
cortex. Subjects performing a word-stem completion task
showed less activation of temporal cortex when they had
previously seen the words before (Squire et al. 1992).
Finally, the fact that neuronal responses decreased with
familiarity suggests a link to orienting responses and behavioral habituation. Although no individual cells in IT cortex
appeared to be novelty detectors, the summed activity of IT
cells could provide a signal to other systems that the current
stimulus is new and deserving of attention. As the organism
orients and attends to the new stimulus, activated IT cell
populations shrink to the critical set necessary for representing the stimulus. This shrinkage reduces the overall activity
in IT cortex, reducing the drive on the orienting system and
freeing the organism’s attention for other, competing, stimuli. One could view this as a negative feedback system for
incorporating knowledge about new stimuli into the structure of the cortex (see Fig. 11). A similar negative feedback
system is an essential component of Carpenter and Grossberg’s ( 1987) adaptive resonance theory (ART) of memory
and attention. Gochin and colleagues (Gochin 1990; Gochin and Lubin 1990; Gochin et al. 199 1) have pointed out
similarities between characteristics of ART and IT neuronal properties and have suggested that ART may serve as a
useful model of memory storage in IT cortex.
Related structures
Anterior-ventral
IT cortex has direct or indirect connections with rhinal cortex, entorhinal cortex, the hippocampal formation, parahippocampal
gyrus, amygdala, the basal
forebrain, and the prefrontal cortex (Aggleton et al. 1980;
Insausti et al. 1987a,b; Turner et al. 1980), all of which
have been implicated in recognition memory. Thus anterior-ventral IT cortex is in a position to supply other structures with information
critical for visual memory formation and to be influenced by those structures in return. On
1928
L. LI,
E. K. MILLER,
the basis of our analysis of average response histograms, it
takes IT cells > 100 ms after response onset to distinguish a
novel sample stimulus from one that has been presented
just once before. This is clearly enough time for the response difference to be caused by feedback from other
structures. However, after just a few repetitions, IT cells
distinguish novel from familiar stimuli within 10 ms of the
onset of the visual response (or GO-90 ms from the onset
of the visual stimulus), which is probably not enough time
for the response difference to be caused by feedback from
memory circuits beyond IT cortex. Thus, although feedback may be necessary to initiate the processes that ultimately result in memory formation in IT, after a few repetitions IT networks appear to be capable of distinguishing
novel from familiar stimuli on their own. It is not known
whether IT cortex is the only structure in the visual system
with this capability. Although neuronal responses are
known to be affected by repeated presentations of a stimulus over very short time intervals in areas posterior to IT
cortex, including V4 (Haenny et al. 1988; Haenny and
Schiller 1988; Maunsell et al. 199 1) and V 1 (Nelson 199 1))
it is not yet known to what extent neurons in these areas can
distinguish familiar stimuli from novel ones.
We thank M. Mishkin
for advice and support throughout
the study, T.
White for writing the computer
programs,
M. Adams and J. Sewell for help
with histology,
J. Hart and R. Hoag for help with the monkeys,
and V.
Brown and P. DeWeerd
for valuable comments
on the manuscript.
The research was supported
in part by the Human
Frontiers
Science
Program Foundation.
Address for reprint requests: R. Desimone,
National
Institute of Mental
Health,
Laboratory
of Neuropsychology,
Building
9, Rm. 1E 104, Bethesda, MD 20892.
Received
13 October
1992; accepted
in final form
29 January
1993.
REFERENCES
J. P., BURTON,
M. J., AND PASSINGHAM,
R. E. Cortical and
afferents to the amygdala
of the rhesus monkey
(Macaca
Bruin Rev. 36: 243-248,
1980.
AMARAL,
D. G., INSAUSTI,
R., AND COWAN,
W. M. The entorhinal
cortex
of the monkey.
I. Cytoarchitectonic
organization.
J. Comp. Neural. 264:
326-355,
1987.
AMARAL,
D. G. AND PRICE, J. L. Amygdalo-cortical
projections
in the
monkey ( Macaca ~fascicularis > . .J. Camp. Neurol. 230: 465-496,
1984.
BAIZER,
J. S., UNGERLEIDER,
L. G., AND DESIMONE,
R. Organization
of
visual inputs to the inferior
temporal
and posterior
par&al
cortex in
macaques. J. Neurosci.
1 1: 168- 190, 199 1.
BAYLIS,
G. C. AND ROLLS,
E. T. Responses of neurons
in the inferior
temporal cortex in short term and serial recognition
memory
tasks. Exp.
Brain Rex 65: 6 14-622,
1987.
CARPENTER,
G. A. AND GROSSBERG,
S. A massively
parallel architecture
for a self-organizing
neural pattern recognition
machine. Comp. Visual
Graphics Image Process. 37: 54- 1 15, 1987.
DESIMONE,
R., ALBRIGHT,
T. D., GROSS, C. G., AND BRUCE, C. Stimulusselective properties
of inferior
temporal
neurons
in the macaque.
J.
Neuro.sci. 4: 205 l-2062,
1984.
DESIMONE,
R., FLEMING,
J., AND GROSS, C. G. Prestriate afferents to inferior temporal
cortex: an HRP study. Brain Res. 184: 4 l-55, 1980.
DESIMONE,
R. AND UNGERLEIDER,
L. G. Neural
mechanisms
of visual
processing in monkeys.
In: Handbook
qf Neuropsychology,
edited by F.
Boller and J. Grafman.
New York: Elsevier,
1989, vol. 2, p. 267-299.
ESKANDAR,
E. N., RICHMOND,
B. J., AND OPTICAN,
L. M. Role of inferior
temporal
neurons in visual memory.
I. Temporal
encoding of information about visual images, recalled images, and behavioral
context.
J.
Neurophysiol.
68: 1277- 1295, 1992.
GAFFAN,
D. AND MURRAY,
E. A. Monkeys
(MacacaTfascicularis)
with
rhinal cortex ablations succeed in object discrimination
learning despite
AC;GLETON,
subcortical
muluttu).
AND
R. DESIMONE
24-hr intertrial
intervals and fail at matching
to sample despite double
sample presentations.
Behuv. Ncurosci.
106: 30-38,
1992.
GOCHIN,
P. M. Pattern recognition
in primate temporal
cortex: but is it
ART? Proc. Int. Joint Conf Neural Networks
1: I77-79,
1990.
GOCHIN,
P. M. AND LUBIN,
J. M. A hierarchical
machine vision system
based on a model of the primate visual system. Proc. IEEE Int. Symp.
Intelligent
Control
1: 6 l-65, 1990.
GOCHIN,
P. M., MILLER,
E. K., GROSS, C. G., AND GERSTEIN,
G. L. Functional interactions
among neurons in macaque inferior temporal
cortex.
Exp. Bruin Rex 84: 505-5 16, 199 1.
GROSS, C. G. Visual functions
of inferotemporal
cortex. In: flandbook CI/
Sensory Physiology,
edited by R. Jung. Berlin: Springer-Verlag,
1973.
GROSS, C. G., BENDER,
D. B., AND GERSTEIN,
G. L. Activity
of inferior
temporal
neurons
in behaving
monkeys.
Nezlropsvchol(~giu
17: 2 15229, 1979.
GROSS, C. G., ROCHA-MIRANDA,
C. E., AND BENDER,
D. B. Visual properties of neurons in inferotemporal
cortex of the macaque.
J. Neurophysiol. 35: 96- 1 1 1, 1972.
HAENNY,
P. E., MAUNSELL.
J. H. R.. AND SCHILLER,
P. H. State dependent
activity
in monkey
visual cortex. II. Retinal and extraretinal
factors in
V4. Exp. Brain Rex 69: 245-259,
1988.
HAENNY,
P. E. AND SCHILLER,
P. H. State dependent
activity in monkey
visual cortex. I. Single cell activity
in V 1 and V4 on visual tasks. Exp.
Brain Rex 69: 225-244,
1988.
HOREL,
J. A., PYTKO-JOINER,
D. E., VOYTKO,
M. L., AND SALSBURY,
K.
The performance
of visual tasks while segments of the inferotemporal
cortex are suppressed by cold. Bchuv. Brain Res. 23: 29-42, 1987.
INSAUSTI,
R., AMARAL,
D. G., AND COWAN,
W. M. The entorhinal
cortex
of the monkey.
III. Subcortical
afferents.
.J. Comp. Neural. 264: 396408, 1987a.
INSAUSTI,
R., AMARAL,
D. G., AND COWAN,
W. M. The entorhinal
cortex
of the monkey.
II. Cortical
afferents.
J. Camp. Neural. 264: 356-395,
1987b.
MARTIN-ELKINS,
C. L. AND HOREL, J. A. Cortical afferents to behaviorally
defined regions of the inferior
temporal
and parahippocampal
gyri as
demonstrated
by WGA-HRP.
.I. Camp. Neural. 32 I : 177- 192, 1992.
MAUNSELL,
J. H. R., SCLAR,
G., NEALEY,
T. A., AND DEPRIEST,
D. D.
Extraretinal
representations
in area V4 in the macaque monkey.
Visuul
Ncurosci.
7: 56 l-573,
199 1.
MEUNIER,
M., MURRAY,
E. A., BACHEVALIER,
J., AND MISHKIN,
M. Effects of perirhinal
cortical lesions on visual recognition
in monkeys.
Sot.
Neurosci.
Ahstr. 16: 6 16, 1990.
MIKAMI,
A. AND KUBOTA,
K. Inferotemporal
neuron activities and color
discrimination
with delay. Bruin Rcs. 182: 65-78,
1980.
MILLER,
E. K., GOCHIN,
P. M.. AND GROSS, C. G. Habituation-like
decrease in the responses of neurons in inferior
temporal
cortex of the
macaque.
Visuul Ncwro.sci. 7: 357-362,
199 1a.
MILLER,
E. K., LI, L., AND DESIMONE,
R. A neural mechanism
for working and recognition
memory
in inferior temporal
cortex. Science Wash.
DC254:
1377-1379,
1991b.
MILLER,
E. K., LI, L., AND DESIMONE,
R. Activity
of neurons in anterior
inferior temporal
cortex during a short-term
memory
task. J. Neuro.sci.
13: 1460-1478,
1993.
MISHKIN,
M. A memory
system in the monkey.
Philos. Trans. R. Sot.
Land. B Biol. Sci. 298: 83-95,
1982.
MURRAY,
E. A. Medial temporal
lobe structures
contributing
to recognition memory:
the amygdaloid
complex
versus the rhinal cortex. In: The
Amygdala:
Neurohiologicul
Aspects of’ Emotion,
Memory,
und Mental
Dy$dnction.
J. P. Aggleton,
New York: Wiley-Liss,
1992, p. 453-470.
MURRAY,
E. A., BACHEVALIER,
J., AND MISHKIN,
M. Effects of rhinal
cortical lesions on visual recognition
memory
in rhesus monkeys.
Sot.
Neurosci. Abstr. 15: 342, 1989.
NELSON,
S. B. Temporal
interactions
in the cat visual system. I. Orientation-selective
suppression
in the visual cortex. J. Neurosci.
1 1: 344-356,
1991.
PHILLIPS,
R. R., MALAMUT,
B. L., BACHEVALIER,
J., AND MISHKIN,
M.
Dissociation
of the effects of inferior
temporal
and limbic lesions on
object discrimination
learning
with 24-h inter-trial
intervals.
Behav.
Brain Rex 27: 99- 107, 1988.
POLLEN,
D. A., NAGLER,
M ., DAUGMAN,
J., KRONA .UER, R., AND CAVANAUGH,
P. Use of Gabor elementary
fu nctions to probe receptive-field
substructure
of posterior
inferotemporal
neurons in the owl monkey.
Vision Rex 24: 233-24 1, 1984.
RICHES,
I. P., WILSON,
F. A., AND BROWN,
M. W. The effects of visual
FAMILIAR
STIMULI
stimulation and memory on neurons of the hippocampal formation and
the neighboring parahippocampal gyrus and inferior temporal cortex of
the primate. J. Neurosci. 11: 1763-1779,
199 1.
RICHMOND,
B. J. AND OPTICAN, L. M. Temporal encoding of two-dimensional patterns by single units in primate inferior temporal cortex. II.
Quantification of response waveform. J. Neurophysiol. 57: 147- 16 1,
1987.
RICHMOND,
B. J., WURTZ, R. H., AND SATO, T. Visual responses of inferior temporal neurons in awake rhesus monkey. J. Neurophysiol. 50:
1415-1432,
1983.
ROLLS, E. T., BAYLIS,
G. C., HASSELMO, M. E., AND NALWA, V. The effect
of learning on the face selective responses of neurons in the cortex in the
superior temporal sulcus of the monkey. Exp. Brain Res. 76: 153- 164,
1989.
SCHACTER,
D. L., COOPER, L. A., AND DELANEY, S. M. Implicit memory
for unfamiliar objects depends on access to structural descriptions. J.
Exp. Psychol. 119: 5-24, 1990.
SCHACTER, D. L., COOPER, L. A., THARAN, M., AND RUBENS, A. B. Preserved priming of novel objects in patients with memory disorders. J.
Cognit. Neurosci. 3: 117- 130, 199 1.
SCHWARTZ, E. L., DESIMONE,
R., ALBRIGHT, T. D., AND GROSS, C. G.
Shape recognition and inferior temporal neurons. Proc. Nat/. Acad. Sci.
USA 80: 5776-5778,
1983.
SHEPARD, R. N. Recognition
memory for words, sentences and pictures. J.
Verb. Learn. Verb. Behav. 6: 156-163, 1992.
SQUIRE, L. R., OJEMANN, J. G., MIEZIN, F. M., PETERSEN, S. E., VIDEEN,
T. O., AND RAICHLE, M. E. Activation of the hippocampus in normal
humans: a functional anatomical study of memory. Proc. Nat/. Acad.
Sci. USA 89: 1837-1841,
1992.
STANDING, L. Learning 10,000 pictures. Q. J. Exp. Psychol. 25: 207-222,
1973.
SUZUKI,
W. A., ZOLA-MORGAN,
S., SQUIRE, L. R., AND AMARAL, D. G.
Lesions of the perirhinal and parahippocampal cortices in monkeys produce a modality general and long lasting memory impairment. Sot.
Neurosci. Abstr. 17: 339, 199 1.
SUZUKI, W. A., ZOLA-MORGAN,
S., SQUIRE, L. R., AND AMARAL, D. G.
AND IT CORTEX
1929
Lesions of the perirhinal and parahippocampal cortices in the monkey
produce long-lasting memory impairment in the visual and tactile modalities. J. Neurosci. In press.
TANAKA,
K., SAITO, H., FUKADA,
Y., AND MORIYA, M. Coding visual
images of objects in the inferotemporal cortex of the macaque monkey.
J. Neurophysiol. 66: 170-l 89, 199 1.
TURNER, B. H., MISHKIN,
M., AND KNAPP, M. Organization of the amygdalopetal projections from modality-specific cortical association areas
in the monkey. J. Comp. Neurol. 19 1: 5 15-543, 1980.
UNGERLEIDER,
L. G. AND MISHKIN, M. Two cortical visual systems. In:
Analysis of Visual Behavior, edited by J. Ingle, M. A. Goodale, and
R. J. W. Mansfield. Cambridge, MA: MIT Press, 1982, p. 549-586.
VAN HOESEN, G. W. AND PANDYA, D. N. Some connections of the entorhinal (area 28) and perirhinal (area 35) cortices of the rhesus monkey.
III. Efferent connections. Brain Res. 95: 48-67, 1975a.
VAN HOESEN, G. W. AND PANDYA, D. N. Some connections of the entorhinal (area 28) and perirhinal (area 35) cortices of the rhesus monkey.
I. Temporal lobe afferents. Brain Res. 95: l-24, 1975b.
VOGELS, R. AND ORBAN, G. A. Effects of task related stimulus attributes
on infero-temporal neurons studied in the discriminating monkey. Sot.
Neurosci. Abstr. 16: 62 1, 1990.
WEBSTER, M. J., UNGERLEIDER,
L. G., AND BACHEVALIER, J. Connections
of inferior temporal areas TE and TEO with medial temporal-lobe structures in infant and adult monkeys. J. Neurosci. 11: 1095- 1116, 199 1.
WILSON, F. A. W. AND ROLLS, E. T. Neuronal responses related to novelty
and familiarity of visual stimuli in the substantia innominata, diagonal
band of Broca and periventricular region of the primate basal forebrain.
Exp. Brain Res. 80: 104- 120, 1990.
ZOLA-MORGAN,
S., SQUIRE, L. R., AMARAL, D. G., AND SUZUKI, W. A.
Lesions of perirhinal and parahippocampal cortex that spare the amygdala and hippocampal formation produce severe memory impairment.
J. Neurosci. 9: 4355-4370, 1989.
ZOLA-MORGAN,
S., SQUIRE, L. R., CLOWER, R. P., AND REMPEL, N. L.
Damage to the perirhinal cortex exacerbates memory impairment following lesions to the hippocampal formation. J. Neurosci. 13: 25 l-265,
1993.